The Digital PDP-1 was about the third computer I ever programmed, after the vacuum-tube Burroughs 220 and the innovative Burroughs B-5500.
The version I used around 1966-1967 - a PDP-1D - was at Stanford's Institute for Mathematical Studies in the Social Sciences, which was developing computer-based instruction programs. It had recently been transferred from John McCarthy's Artificial Intelligence Lab, which had just received a PDP-6. (I was developing programs for teaching high-school geometry).
The PDP-1 had 18-bit words, used 1's complement arithmetic (octal notation was used, 1960 - the year of the first PDP-1's development - was before IBM invented the word 'byte' and popularized hexidecimal notation), and used asynchronous logic - using time delays instead of a master clock. The PDP-1D differed from the more common version by having an asynchronous memory bus with handshaking like the PDP-6 and the later PDP-10. Logic circuits were negative-level DTL (diode-transistor logic) with discrete germanium transistors and diodes. Using germanium instead of silicon transistors made the machine slightly faster with then-current technology, but meant that temperature control was essential, since the conductivity of germanium changes rapidly with temperature. The PDP-6 also used germanium transistors and diodes - in fact the same logic modules.
In fact, Digital started out making logic modules a bit before they decided to build the PDP-1. This was an important part of their business for many years. Their principal competitor in this area was SDS (Scientific Data Systems) in Southern California - which built the competing Sigma series of computers. SDS was later purchased by Xerox, and renamed XDS. The SDS-940 played an important part in early developments at Doug Englebart's Augmentation Research Center at SRI, where the mouse and hypertext first appeared. The SRI 940 was replaced by a specially-modified PDP-10 around 1970 - the modifications, made my BB&N, added paged memory and helped make this machine one of the first few hosts on the ARPA Net. (I worked at this time for a contractor which built interfaces for the 940-to-PDP-10 conversion, and a friend who worked for ARC wrote the code which sent the first packet on the ARPA Net).
The only other PDP-1D model I know about was at Bolt, Beranack and Newman, later the prime contractors for the ARPA Net.
The saga of Xerox PARC's refusing to by a Xerox SIGMA-9 and building MAXX, their own clone of a PDP-10, is another story, but one integral in the development of GUI computing at Xerox, since memory components designed for MAXX were used in the Alto.